Credit ratings cliff

Why does one notch of downgrade cost a borrower far more than the notch before it?

Credit ratings run down a long ladder of notches. A borrower slips from A to A minus, then to BBB, then BBB minus. Each step is a small revision of opinion, and each costs a little more to borrow.

Then comes one more step — to BB plus — and the cost does not rise a little. Access to whole classes of buyer disappears, the bond's price gaps, covenants may bite, and in bad cases collateral has to be posted the same week. Nothing about the borrower changed more between those two notches than between the previous two. So what, exactly, is different about that particular line?

Start by asking what a rating actually is. It is an opinion about the likelihood of not being repaid, expressed on an ordered scale. The underlying quantity it tracks — default probability — is continuous. It does not leap between BBB minus and BB plus; the ladder is a coarse rendering of something smooth.

So if the thing being measured is continuous, a discontinuity in consequences cannot be coming from the borrower. It has to be coming from something that reads the measurement. Where would you look for it?

Look at the rulebooks. An enormous number of them draw a single binary distinction at exactly that point: investment grade above, speculative below. Bond indices define their membership there, so an index-tracking fund must sell a bond that crosses it — not because its manager formed a view, but because the benchmark no longer contains it. Investment mandates for pension funds and insurers are written the same way. Bank and insurance capital rules charge more against the lower bucket. Central banks and clearing houses set collateral eligibility by it. Individual loan agreements and derivative contracts embed rating triggers that fire on the crossing.

Now count what happens simultaneously when a borrower crosses. A large, largely price-insensitive set of holders becomes obliged to sell at once. The natural buyers are a different, smaller pool with its own capital constraints. Supply arrives without demand, so the price falls further than the change in opinion warrants — and note that the sellers are not disagreeing about value. They are complying.

Then follow the consequence round. A wider spread is a higher cost of funding. A higher cost of funding weakens the borrower. A weaker borrower is more likely to be downgraded again. The threshold, having been crossed, makes further crossing more likely — so the discontinuity is not merely a cost, it is partly self-confirming.

There is a second effect, and it is visible on the other side of the line. If a threshold is expensive, borrowers work hard to stay above it. Companies sitting one notch clear will sell assets, cut dividends or postpone acquisitions specifically to defend the rating — steps they would not take for a downgrade from A to A minus. That produces a crowding of issuers just above the boundary, and it means the threshold shapes real corporate behaviour years before anyone approaches it.

Notice what has been explained and what has not. Nothing here required the rating agencies to be wrong or the boundary to be badly placed. The cliff is manufactured entirely by the fact that many independent parties wrote the same discrete rule against the same published label. Any threshold used widely enough as a coordination device acquires this property, whatever it happens to measure.

The default probabilities do differ across the line, so the boundary is not arbitrary. Historically, speculative-grade default rates are substantially higher than investment-grade ones, and averaged over many years the gap between BBB and BB cohorts is real. The claim is not that the distinction is meaningless; it is that the jump in consequences is far larger than the jump in underlying risk at that one adjacent pair of notches.

Forced selling is not the same as informed selling, and this matters for what the price means. When a bond gaps on crossing, part of the move is news about the borrower and part is a queue of mandated sellers. Studies of fallen angels find a characteristic pattern of price pressure around the downgrade followed by partial recovery, which is what a mechanical-selling account predicts and a pure-information account does not.

Rating triggers are the sharpest version and the least visible. A contract clause requiring extra collateral on a downgrade converts an opinion into an immediate cash demand. This is the mechanism that turned a downgrade into an acute liquidity crisis for AIG in 2008, and it remains the reason treasurers care about a notch far more than the spread arithmetic alone would justify.

Reforms aimed at removing hard-wired ratings have had limited effect. Regulators after 2008 pushed to strip mechanical rating references out of rules, and some were removed. But private contracts and index definitions were not reachable by that, so the coordination point survived. That is instructive: the cliff lives in the collective use of the label, not in any single rule that can be repealed.

The falsification test. If the cliff comes from rules keyed to the label rather than from the credit itself, then a borrower whose fundamentals deteriorate by the same amount without crossing the line should not see the same jump. That is broadly what is observed — the notch that matters is the boundary one, not the equally sized step above it.

• Whether index providers could soften the cliff with gradual weighting rather than binary inclusion.
• How sovereign downgrades transmit the same effect to every borrower under the sovereign ceiling.
• The behaviour of BBB minus issuers in a downturn, and whether defensive action actually works.

Every term the collection defines is gathered in the glossary.